Fascinating look at how prediction markets are calibrating risk assessments in real time. The jump from 5% to 30% probability after Venezuela shows how quickly geopolitical norms can shift when rhetoric gets backed by action. That conditional removal probabiliy is really telling about institutinal checks.
Apparently "Ben the bullshit detector" doesn't know what the little man icon means in these manifold trade screenshots - It's the number of traders Ben! Most of the "data" for this article is based on Manifold bets with less than 30 traders i.e. it is meaningless.
You can get a pretty good aggregate forecast with just a handful of forecasters. Same with markets. I suggest reading the literature on wisdom of crowds; you’d be surprised. Manifold’s calibration on markets with 20-30 traders is quite good. Diminishing returns with higher numbers of traders.
And yet I feel certain that you would never have written the article with polls of 30 of your friends in place of the Manifold screen shots even though you seem to believe that that would have been equally valid. A small number of forecasters can be valuable, but only when they are known to be informed forecasters - You have no basis for assuming this. Also Manifold is a betting platform as much as a forecasting platform. They could just be gambling addicts who liked the odds - I've placed many Manifold bets that I don't much believe in just because the odds seem good.
If you feel that the markets in this article are mispriced, feel free to bet on them and make some mana :)
This is a Manifold newsletter. We’re limited by the availability of Manifold markets on complex, fast-moving events. One of the advantages of this platform is that users can make their own markets, asking questions with more complicated/subjective/niche topics and resolution criteria than larger, real-money platforms. Of course there are tradeoffs (smaller volume like you’ve said). But I think 10 nuanced markets on breaking news with 30 traders each can often provide far more information than a single market with 1000 traders.
As I said, I DO bet on mispriced Manifold markets.
Regarding the bit about 10 markets x 30 traders : That would be true IF they were independent but, for all we know, the same 30 traders are making the same bet/forecast in each market. The reason I picked 30 traders is that does appear to be the approximate number on many of your markets in the article which does tend to support the idea that they contain many of the same traders and therefore it is not unsurprising that they appear strongly correlated.
That's true, they could be correlated. But also, about half the markets in this article have more in the range of 80-150 traders, which I think is absolutely large enough to pretty a pretty meaningful signal. The most load-bearing market (will Trump buy/acquire part of Greenland) has over 1000 traders. One purpose of this newsletter is to get people trading on interesting markets with low trading volume! So I always like to include a few smaller markets with only a handful of traders.
And by the way, yes, I’ve ran forecasting tournaments for my non-forecasting friends. And you’d be surprised how good the crowd aggregate does in these tournaments! A handful of amateurs can beat seasoned superforecasters. Markets provide even more signal on top of a simple average, by weighting for conviction and prior performance, although they have their own set of limitations.
I think you identified the problem there in that you separated forecasting and markets. Manifold combines both and that causes unquantifiable interactions.
I live in Rockville, a small city in Maryland, that has a population a decent bit larger than the entire 2.166 million sq km of Greenland. When the Inuit arrived in Greenland around 1000 AD, a few decades after the voyages of Erik the Red, and displaced the local population, it was a barren wasteland, as it remains today.
Fascinating look at how prediction markets are calibrating risk assessments in real time. The jump from 5% to 30% probability after Venezuela shows how quickly geopolitical norms can shift when rhetoric gets backed by action. That conditional removal probabiliy is really telling about institutinal checks.
Some Fruits Are Orange
Apparently "Ben the bullshit detector" doesn't know what the little man icon means in these manifold trade screenshots - It's the number of traders Ben! Most of the "data" for this article is based on Manifold bets with less than 30 traders i.e. it is meaningless.
You can get a pretty good aggregate forecast with just a handful of forecasters. Same with markets. I suggest reading the literature on wisdom of crowds; you’d be surprised. Manifold’s calibration on markets with 20-30 traders is quite good. Diminishing returns with higher numbers of traders.
And yet I feel certain that you would never have written the article with polls of 30 of your friends in place of the Manifold screen shots even though you seem to believe that that would have been equally valid. A small number of forecasters can be valuable, but only when they are known to be informed forecasters - You have no basis for assuming this. Also Manifold is a betting platform as much as a forecasting platform. They could just be gambling addicts who liked the odds - I've placed many Manifold bets that I don't much believe in just because the odds seem good.
If you feel that the markets in this article are mispriced, feel free to bet on them and make some mana :)
This is a Manifold newsletter. We’re limited by the availability of Manifold markets on complex, fast-moving events. One of the advantages of this platform is that users can make their own markets, asking questions with more complicated/subjective/niche topics and resolution criteria than larger, real-money platforms. Of course there are tradeoffs (smaller volume like you’ve said). But I think 10 nuanced markets on breaking news with 30 traders each can often provide far more information than a single market with 1000 traders.
As I said, I DO bet on mispriced Manifold markets.
Regarding the bit about 10 markets x 30 traders : That would be true IF they were independent but, for all we know, the same 30 traders are making the same bet/forecast in each market. The reason I picked 30 traders is that does appear to be the approximate number on many of your markets in the article which does tend to support the idea that they contain many of the same traders and therefore it is not unsurprising that they appear strongly correlated.
That's true, they could be correlated. But also, about half the markets in this article have more in the range of 80-150 traders, which I think is absolutely large enough to pretty a pretty meaningful signal. The most load-bearing market (will Trump buy/acquire part of Greenland) has over 1000 traders. One purpose of this newsletter is to get people trading on interesting markets with low trading volume! So I always like to include a few smaller markets with only a handful of traders.
And by the way, yes, I’ve ran forecasting tournaments for my non-forecasting friends. And you’d be surprised how good the crowd aggregate does in these tournaments! A handful of amateurs can beat seasoned superforecasters. Markets provide even more signal on top of a simple average, by weighting for conviction and prior performance, although they have their own set of limitations.
I think you identified the problem there in that you separated forecasting and markets. Manifold combines both and that causes unquantifiable interactions.
Seems a little harsh to call the home of the Inuits a barren wasteland.
I live in Rockville, a small city in Maryland, that has a population a decent bit larger than the entire 2.166 million sq km of Greenland. When the Inuit arrived in Greenland around 1000 AD, a few decades after the voyages of Erik the Red, and displaced the local population, it was a barren wasteland, as it remains today.